Fix/task publish estimation procedure

[RugvedBane]

Fix

Closes #1751

After create_task() + publish(), the task object had estimation_procedure = None
and no data_splits_url, making get_split_dimensions() fail.

Root cause

OpenMLTask._parse_publish_response() only assigned task_id from the server
response. The fields estimation_procedure_type, estimation_parameters, and
data_splits_url were never populated, because the publish endpoint does not
return them.

Fix

After setting task_id in _parse_publish_response(), immediately fetch the
complete task from the server using _get_task_description(task_id) and copy
estimation_procedure and estimation_procedure_id onto self.

The import of _get_task_description is kept local to avoid a circular import
(functions.py imports from task.py at the top level).

Changes

• openml/tasks/task.py — 3 lines added to _parse_publish_response()
• tests/test_tasks/test_task_functions.py — regression test added
• tests/files/mock_responses/tasks/task_upload_successful.xml — mock fixture

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 81.36%. Comparing base (4a25027) to head (f26221b).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1756      +/-   ##
==========================================
- Coverage   81.50%   81.36%   -0.14%     
==========================================
  Files          63       63              
  Lines        5238     5242       +4     
==========================================
- Hits         4269     4265       -4     
- Misses        969      977       +8     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[RugvedBane]

Hi @L-Agbaria, just wanted to check in. The 3 failing checks appear to be pre-existing failures unrelated to this PR. I checked two other PRs opened around the same time and they have the same 3 failures on the same Python/sklearn matrix combinations. Happy to investigate further if needed, but wanted to flag this in case it is blocking the review. Let me know if there is anything else you would like me to address!

[L-Agbaria]

Hello @RugvedBane, I will look into your implementation soon. The test fail due to the rate limit issue #1757, this will be addressed later and should not delay this PR.
I'm still new to the OpenML project and am therefore a bit distracted by multiple things at the same time, reviewing your implementation is high on my agenda and I will get back to you as soon as possible.

[RugvedBane]

No issue at all, I just wanted to make sure there was nothing blocking from my side. Take your time!